Kimi K3

NewKimi

Ranks #3 of 43 on reading real contracts into structured billing data — +4.4 pts vs GPT-5.5.

73.7%

Accuracy

#3 of 43

8.7%

Hallucination

HIGH-confidence & wrong

$278

Cost / 1k contracts

$0.28 each · via OpenRouter

180.6s

Median latency

p90 397.4s

±3

Run-to-run σ

3 runs

0.26

Value

acc. pts per $/1k

Tokens per contract

Input21,517

Output14,243

Reasoning10,102

Reliability 100% · valid structured output across 18 calls.

Want the full picture? How we score →